drm/amdgpu: reset vm state machine after gpu reset(vram lost)
[Why]
Page table of compute VM in the VRAM will lost after gpu reset.
VRAM won't be restored since compute VM has no shadows.
[How]
Use higher 32-bit of vm->generation to record a vram_lost_counter.
Reset the VM state machine when vm->genertaion is not equal to
the new generation token.
v2: Check vm->generation instead of calling drm_sched_entity_error
in amdgpu_vm_validate.
v3: Use new generation token instead of vram_lost_counter for check.
The ras command shared memory is allocated from
VRAM and the response status of the command
buffer will not be zero due to gpu being in
fatal error state after ras UE error injection.
drm/amd/display: Remove ASSERT if significance is zero in math_ceil2
In the DML math_ceil2 function, there is one ASSERT if the significance
is equal to zero. However, significance might be equal to zero
sometimes, and this is not an issue for a ceil function, but the current
ASSERT will trigger warnings in those cases. This commit removes the
ASSERT if the significance is equal to zero to avoid unnecessary noise.
Alex Deucher [Tue, 9 Jul 2024 21:54:11 +0000 (17:54 -0400)]
drm/amdgpu/sdma5.2: Update wptr registers as well as doorbell
We seem to have a case where SDMA will sometimes miss a doorbell
if GFX is entering the powergating state when the doorbell comes in.
To workaround this, we can update the wptr via MMIO, however,
this is only safe because we disallow gfxoff in begin_ring() for
SDMA 5.2 and then allow it again in end_ring().
Enable this workaround while we are root causing the issue with
the HW team.
drm/i915/gt: Do not consider preemption during execlists_dequeue for gen8
We're seeing a GPU hang issue on a CHV platform, which was caused by commit bac24f59f454 ("drm/i915/execlists: Enable coarse preemption boundaries for
Gen8").
The Gen8 platform only supports timeslicing and doesn't have a preemption
mechanism, as its engines do not have a preemption timer.
Commit 751f82b353a6 ("drm/i915/gt: Only disable preemption on Gen8 render
engines") addressed this issue only for render engines. This patch extends
that fix by ensuring that preemption is not considered for all engines on
Gen8 platforms.
v4:
- Use the correct Fixes tag (Rodrigo Vivi)
- Reworded commit log (Andi Shyti)
v3:
- Inside need_preempt(), condition of can_preempt() is not required
as simplified can_preempt() is enough. (Chris Wilson)
v2: Simplify can_preempt() function (Tvrtko Ursulin)
The Samsung ATNA45AF01 panel is an AMOLED eDP panel that has backlight
control over the DP AUX channel. While it works almost correctly with the
generic "edp-panel" compatible, the backlight needs special handling to
work correctly. It is similar to the existing ATNA33XC20 panel, just with
a larger resolution and size.
Add a new "samsung,atna45af01" compatible to describe this panel in the DT.
Use the existing "samsung,atna33xc20" as fallback compatible since existing
drivers should work as-is, given that resolution and size are discoverable
through the eDP link.
Matthew Brost [Tue, 16 Jul 2024 06:39:01 +0000 (23:39 -0700)]
drm/xe: Wedge the entire device
Wedge the entire device, not just GT which may have triggered the wedge.
To implement this, cleanup the layering so xe_device_declare_wedged()
calls into the lower layers (GT) to ensure entire device is wedged.
While we are here, also signal any pending GT TLB invalidations upon
wedging device.
Lastly, short circuit reset wait if device is wedged.
v2:
- Short circuit reset wait if device is wedged (Local testing)
Michal Wajdeczko [Thu, 11 Jul 2024 19:23:19 +0000 (21:23 +0200)]
drm/xe/pf: Limit fair VF LMEM provisioning
Due to the current design of the BO and VRAM manager, any object
with XE_BO_FLAG_PINNED flag, which the PF driver uses during VF
LMEM provisionining, is created with the TTM_PL_FLAG_CONTIGUOUS
flag, which may cause VRAM fragmentation that prevents subsequent
allocations of larger objects, like fair VF LMEM provisioning.
To avoid such failures, round down fair VF LMEM provisioning size
to next power of two size, to compensate what xe_ttm_vram_mgr is
doing to achieve contiguous allocations.
Documentation/gpu/amdgpu/display/display-manager:134:
drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h:3: WARNING: Duplicate C
declaration, also defined at gpu/amdgpu/display/dcn-blocks:101.
Declaration is '.. c:struct:: mpcc_blnd_cfg'.
Documentation/gpu/amdgpu/display/display-manager:146:
drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h:3: WARNING: Duplicate C
declaration, also defined at gpu/amdgpu/display/dcn-blocks:3.
Declaration is '.. c:enum:: mpcc_alpha_blend_mode'.
To address the above warnings, this commit uses the 'no-identifiers'
option in the dcn-blocks to avoid duplication with the previous use of
this function doc in the display-manager file. Finally, replaces the
deprecated ':function:' in favor of ':identifiers:'.
Documentation/gpu: Remove undocumented files from dcn-blockshubbub.h
The dchubbub.h and hubp.h do not have any meaningful documentation; for
this reason, this commit removes those files from the dcn-blocks
documentation.
drm/amd/display: Add simple struct doc to remove doc build warning
This commit is a part of a series that addresses the following build
warning for opp:
./drivers/gpu/drm/amd/display/dc/inc/hw/opp.h:1: warning: no structured
comments found
./drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h:1: warning: no structured
comments found
This commit fixes this issue by adding a simple kernel-doc to a struct
in the opp.h and the dpp.h files.
When building the kernel-doc, it has the following complaints:
Documentation/gpu/amdgpu/display/dcn-blocks:23:
drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h:3: WARNING: Duplicate C
declaration, also defined at gpu/amdgpu/display/dcn-blocks:3.
Declaration is '.. c:struct:: surface_flip_registers'.
Documentation/gpu/amdgpu/display/dcn-blocks:35:
drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h:3: WARNING: Duplicate C
declaration, also defined at gpu/amdgpu/display/dcn-blocks:3.
Declaration is '.. c:struct:: surface_flip_registers'.
This error happened due to a copy-and-paste where the same file path was
duplicated multiple times to a different set of blocks. This commit
addresses this issue by using the correct file path.
Documentation/gpu: Remove ':export:' option from DCN documentation
This commit reduces, but does not fix, all the occurrences and some of
the documentation warnings related to the 'no structured comments.' This
was caused by the wrong use of the ':export:' option in the DCN
kernel-doc, so this commit drops the usage of those options.
drm/amd/display: Move DIO documentation to the right place
When building the kernel-doc, it complains with the below warning:
./drivers/gpu/drm/amd/display/dc/link/hwss/link_hwss_dio.h:1: warning: no structured comments found
./drivers/gpu/drm/amd/display/dc/link/hwss/link_hwss_dio.h:1: warning: no structured comments found
This warning was caused by the wrong use of the ':export:' and the lack
of function documentation in the file pointed under the ':internal:'.
This commit addresses those issues by relocating the overview
documentation to the correct C file, removing the ':export:' options,
and adding two simple kernel-doc to ensure that ':internal:' does not
have any warning.
Tim Huang [Fri, 12 Jul 2024 03:05:07 +0000 (11:05 +0800)]
drm/amd/pm: early return if disabling DPMS for GFX IP v11.5.2
This was intended to add support for GFX IP v11.5.2, but it needs
to be applied to all GFX11 and subsequent APUs. Therefore the code
should be revised to accommodate this.
Roman Li [Mon, 15 Jul 2024 20:45:46 +0000 (16:45 -0400)]
drm/amd/display: Add function banner for idle_workqueue
[Why]
htmldocs warning:
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h: warning:
Function parameter or struct member 'idle_workqueue' not described in
'amdgpu_display_manager'.
[How]
Add comment section for idle_workqueue with param description.
Imre Deak [Mon, 8 Jul 2024 19:00:25 +0000 (22:00 +0300)]
drm/i915/dp: Don't switch the LTTPR mode on an active link
Switching to transparent mode leads to a loss of link synchronization,
so prevent doing this on an active link. This happened at least on an
Intel N100 system / DELL UD22 dock, the LTTPR residing either on the
host or the dock. To fix the issue, keep the current mode on an active
link, adjusting the LTTPR count accordingly (resetting it to 0 in
transparent mode).
v2: Adjust code comment during link training about reiniting the LTTPRs.
(Ville)
Imre Deak [Mon, 8 Jul 2024 19:00:24 +0000 (22:00 +0300)]
drm/i915/dp: Reset intel_dp->link_trained before retraining the link
Regularly retraining a link during an atomic commit happens with the
given pipe/link already disabled and hence intel_dp->link_trained being
false. Ensure this also for retraining a DP SST link via direct calls to
the link training functions (vs. an actual commit as for DP MST). So far
nothing depended on this, however the next patch will depend on
link_trained==false for changing the LTTPR mode to non-transparent.
Increase the KMS minor version to indicate GFX12 DCC support since this
contains a major change in how DCC is managed across IPs like GFX, DCN
etc. This will be used mainly by userspace like Mesa to figure out
DCC support on GFX12 hardware.
drm/v3d: Fix Indirect Dispatch configuration for V3D 7.1.6 and later
`args->cfg[4]` is configured in Indirect Dispatch using the number of
batches. Currently, for all V3D tech versions, `args->cfg[4]` equals the
number of batches subtracted by 1. But, for V3D 7.1.6 and later, we must not
subtract 1 from the number of batches.
Implement the fix by checking the V3D tech version and revision.
Fixes several `dEQP-VK.synchronization*` CTS tests related to Indirect Dispatch.
drm/omap: Restrict compile testing to PAGE_SIZE less than 64KB
Prior to commit dc6fcaaba5a5 ("drm/omap: Allow build with
COMPILE_TEST=y"), it was only possible to build the omapdrm driver with
a 4KB page size. After that change, when the PAGE_SIZE is 64KB or
larger, clang points out that the driver has some assumptions around the
page size implicitly by passing PAGE_SIZE to a parameter with a type of
u16:
drivers/gpu/drm/omapdrm/omap_gem.c:758:7: error: implicit conversion from 'unsigned long' to 'u16' (aka 'unsigned short') changes value from 65536 to 0 [-Werror,-Wconstant-conversion]
757 | block = tiler_reserve_2d(fmt, omap_obj->width, omap_obj->height,
| ~~~~~~~~~~~~~~~~
758 | PAGE_SIZE);
| ^~~~~~~~~
arch/powerpc/include/asm/page.h:25:34: note: expanded from macro 'PAGE_SIZE'
25 | #define PAGE_SIZE (ASM_CONST(1) << PAGE_SHIFT)
| ~~~~~~~~~~~~~^~~~~~~~~~~~~
drivers/gpu/drm/omapdrm/omap_gem.c:1504:44: error: implicit conversion from 'unsigned long' to 'u16' (aka 'unsigned short') changes value from 65536 to 0 [-Werror,-Wconstant-conversion]
1504 | block = tiler_reserve_2d(fmts[i], w, h, PAGE_SIZE);
| ~~~~~~~~~~~~~~~~ ^~~~~~~~~
arch/powerpc/include/asm/page.h:25:34: note: expanded from macro 'PAGE_SIZE'
25 | #define PAGE_SIZE (ASM_CONST(1) << PAGE_SHIFT)
| ~~~~~~~~~~~~~^~~~~~~~~~~~~
2 errors generated.
As there is a lot of use of a u16 type throughout this driver and it
will only ever be run on hardware that has a 4KB page size, just
restrict compile testing to when the page size is less than 64KB (as no
other issues have been discussed and it keeps compile testing relatively
more available).
Dave Airlie [Fri, 12 Jul 2024 00:42:16 +0000 (10:42 +1000)]
Merge tag 'drm-misc-next-fixes-2024-07-11' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next
A fix for fbdev on big endian systems, a condition fix for a sharp panel
at removal, and a fix for qxl to prevent unpinned buffer access under
certain conditions.
Matthew Brost [Mon, 8 Jul 2024 21:10:08 +0000 (14:10 -0700)]
drm/xe: Drop trace_xe_hw_fence_free
fence->ctx may be stale memory when trace_xe_hw_fence_free is called
resuling UAF bug when deriving the device name. This tracepoint is not
all that useful, so just drop it.
Ashutosh Dixit [Wed, 3 Jul 2024 16:48:01 +0000 (09:48 -0700)]
drm/xe/uapi: Rename xe perf layer as xe observation layer
In Xe, the perf layer allows capture of HW counter streams. These HW
counters are generally performance related but don't have to be necessarily
so. Also, the name "perf" is a carryover from i915 and is not preferred.
Here we propose the name "observation" for this common layer which allows
capture of different types of these counter streams.
v2: Rename observability layer to observation layer (Lucas/Rodrigo)
v3: Rename sysctl file to "observation_paranoid" (Jose)
drm/amdgpu: timely save bad pages to eeprom after gpu ras reset is completed
The problem case is as follows:
1. GPU A triggers a gpu ras reset, and GPU A drives
GPU B to also perform a gpu ras reset.
2. After gpu B ras reset started, gpu B queried a DE
data. Since the DE data was queried in the ras reset
thread instead of the page retirement thread, bad
page retirement work would not be triggered. Then
even if all gpu resets are completed, the bad pages
will be cached in RAM until GPU B's bad page retirement
work is triggered again and then saved to eeprom.
This patch can save the bad pages to eeprom in time after gpu
ras reset is completed.
v2:
1. Add the above description to code comments.
2. Reuse existing function.
drm/amd/pm: Ignore initial value in smu response register
Why:
If the reg mmMP1_SMN_C2PMSG_90 is being written to during amdgpu driver
load or driver unload, subsequent amdgpu driver load will fail at
smu_hw_init. The default of mmMP1_SMN_C2PMSG_90 register at a clean
environment is 0x1 and if value differs from expected, amdgpu driver
load will fail.
How to fix:
Ignore the initial value in smu response register before the first smu
message is sent,if smc in SMU_FW_INIT state, just proceed further to
send the message. If register holds an unexpected value after smu message
was sent set, smc_state to SMU_FW_HANG state and no further smu messages
will be sent.
v2:
Set SMU_FW_INIT state at the start of smu hw_init/resume.
Check smc_fw_state before sending smu message if in hang state skip
sending message.
Set SMU_FW_HANG only in case unexpected value is detected
For SOCs with GFX v9.4.3, a VF may have multiple compute partitions.
Fetch the partition information during init and initialize partition
nodes. There is no support to switch partition mode in VF mode, hence
disable the same.
Gavin Wan [Mon, 8 Jul 2024 17:07:04 +0000 (17:07 +0000)]
drm/amd/amdgpu: fix SDMA IRQ client ID <-> req mapping.
sdma has 2 instances in SRIOV cpx mode. Odd numbered VFs have
sdma0/sdma1 instances. Even numbered vfs have sdma2/sdma3. For
Even numbered vfs, the sdma2 & sdma3 (irq srouce id
CLIENTID_SDMA2 and CLIENTID_SDMA3) should map to irq seq 0 & 1.
Add qxl_bo_pin_and_vmap() that pins and vmaps a buffer object in one
step. Update callers of the regular qxl_bo_vmap(). Fixes a bug where
qxl accesses an unpinned buffer object while it is being moved; such
as with the monitor-description BO. An typical error is shown below.
[ 4.303586] [drm:drm_atomic_helper_commit_planes] *ERROR* head 1 wrong: 65376256x16777216+0+0
[ 4.586883] [drm:drm_atomic_helper_commit_planes] *ERROR* head 1 wrong: 65376256x16777216+0+0
[ 4.904036] [drm:drm_atomic_helper_commit_planes] *ERROR* head 1 wrong: 65335296x16777216+0+0
[ 5.374347] [drm:qxl_release_from_id_locked] *ERROR* failed to find id in release_idr
Commit b33651a5c98d ("drm/qxl: Do not pin buffer objects for vmap")
removed the implicit pin operation from qxl's vmap code. This is the
correct behavior for GEM and PRIME interfaces, but the pin is still
needed for qxl internal operation.
Also add a corresponding function qxl_bo_vunmap_and_unpin() and remove
the old qxl_bo_vmap() helpers.
Future directions: BOs should not be pinned or vmapped unnecessarily.
The pin-and-vmap operation should be removed from the driver and a
temporary mapping should be established with a vmap_local-like helper.
See the client helper drm_client_buffer_vmap_local() for semantics.
v2:
- unreserve BO on errors in qxl_bo_pin_and_vmap() (Dmitry)
drm/panel: sharp-lq101r1sx01: Fixed reversed "if" in remove
Commit d7d473d8464e ("drm/panel: sharp-lq101r1sx01: Don't call disable
at shutdown/remove") had a subtle bug. We should be calling
sharp_panel_del() when the "sharp" variable is non-NULL, not when it's
NULL. Fix.
Kent Russell [Thu, 4 Jul 2024 14:51:44 +0000 (10:51 -0400)]
Documentation/amdgpu: Clarify MI200 and MI300 entries
Add "Series" to MI200 and MI300 to clarify that they represent the
series of cards, and to more closely match the product information
materials. This also matches other entries in this list
Also correct a typo in the MI300 codename (Vangaram->Vanjaram)
Wayne Lin [Thu, 23 May 2024 04:18:07 +0000 (12:18 +0800)]
drm/amd/display: Solve mst monitors blank out problem after resume
[Why]
In dm resume, we firstly restore dc state and do the mst resume for topology
probing thereafter. If we change dpcd DP_MSTM_CTRL value after LT in mst reume,
it will cause light up problem on the hub.
[How]
Revert commit 202dc359adda ("drm/amd/display: Defer handling mst up request in resume").
And adjust the reason to trigger dc_link_detect by DETECT_REASON_RESUMEFROMS3S4.
Yang Wang [Thu, 4 Jul 2024 05:48:19 +0000 (13:48 +0800)]
drm/amdgpu: remove redundant semicolons in RAS_EVENT_LOG
remove redundant semicolons in RAS_EVENT_LOG to avoid
code format check warning.
Fixes: b712d7c20133 ("drm/amdgpu: fix compiler 'side-effect' check issue for RAS_EVENT_LOG()") Signed-off-by: Yang Wang <[email protected]> Reviewed-by: Tao Zhou <[email protected]> Signed-off-by: Alex Deucher <[email protected]>
Yifan Zha [Thu, 27 Jun 2024 07:06:23 +0000 (15:06 +0800)]
drm/amdgpu: Set no_hw_access when VF request full GPU fails
[Why]
If VF request full GPU access and the request failed,
the VF driver can get stuck accessing registers for an extended period during
the unload of KMS.
[How]
Set no_hw_access flag when VF request for full GPU access fails
This prevents further hardware access attempts, avoiding the prolonged
stuck state.
Thomas Huth [Tue, 2 Jul 2024 12:17:37 +0000 (14:17 +0200)]
drm/fbdev-dma: Fix framebuffer mode for big endian devices
The drm_mode_legacy_fb_format() function only generates formats suitable
for little endian devices. switch to drm_driver_legacy_fb_format() here
instead to take the device endianness into consideration, too.
Daniel Vetter [Fri, 5 Jul 2024 10:45:40 +0000 (12:45 +0200)]
Merge tag 'drm-msm-next-2024-07-04' of https://gitlab.freedesktop.org/drm/msm into drm-next
Updates for v6.11
Core:
- SM7150 support
DPU:
- SM7150 support
- Fix DSC support for DSI panels in video mode
- Fixed TE vsync source support for DSI command-mode panels
- Fix for devices without UBWC in the display controller (ie.
QCM2290)
DSI:
- Remove unused register-writing wrappers
- Fix DSC support for panels in video mode
- Add support for parsing TE vsync source
- Add support for MSM8937 (28nm DSI PHY)
MDP5:
- Add support for MSM8937
- Fix configuration for MSM8953
GPU:
- Split giant device table into per-gen "hw catalog" similar to
what is done on the display side of the driver
- Fix a702 UBWC mode
- Fix unused variably warnings
- GPU memory traces
- Add param for userspace to know if raytracing is supported
- Memory barrier cleanup and GBIF unhalt fix
- X185 support (aka gpu in X1 laptop chips)
- a505 support
- fixes
Daniel Vetter [Fri, 5 Jul 2024 10:37:21 +0000 (12:37 +0200)]
Merge tag 'drm-misc-next-2024-07-04' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next
drm-misc-next for $kernel-version:
UAPI Changes:
Cross-subsystem Changes:
Core Changes:
- dp/mst: Fix daisy-chaining at resume
- dsc: Add helper to dump the DSC configuration
- tests: Add tests for the new monochrome TV mode variant
projects / linux.git / log